Data Driven Approach to Thermal Comfort Model Design

Apart from the dominant environmental factors such as relative humidity, radiant, and ambient temperatures, studies have confirmed that thermal comfort significantly depends on internal personal parameters such as metabolic rate, age, and health status. This study reviews the sensitivity of the Predicted Mean Vote (PMV) thermal comfort model relative to its environmental and personal parameters of a group of people in a space. PMV model equations adapted in ASHRAE Standard 55–Thermal Environmental Conditions for Human Occupancy, are used in this investigation to conduct a parametric study by generating and analyzing multi-dimensional comfort zone plots. It has been found that personal parameters such as metabolic rate and clothing have the highest impact. Current and newly emerging advancements in state of the art wearable technology have made it possible to continuously acquired biometric information. This work proposes to access and exploit this data to build a new innovative thermal comfort model. Relying on various supervised machine-learning methods, a thermal comfort model has been produced and compared to a general model to show its superior performance. Finally, the study represents an architecture to employ new thermal comfort model in inexpensive, responsive and extensible smart home service. Advisor: Fadi Alsalee